The rapid and widespread transmission of COVID-19 has necessitated the development of efficient diagnostic tools. While RT-PCR remains the standard method for diagnosis, its limitations in terms of time and resource intensity highlight the need for alternative solutions. This study addresses this gap by proposing a three-stage hybrid methodology for the rapid identification of COVID-19 using CT scans. In the first stage, pre-trained convolutional neural networks (CNNs), including Vgg-16, ResNet50, and MobileNet-v2, are utilized to extract relevant features from COVID-19-affected lungs. The second stage enhances feature selection through the application of meta-heuristic techniques such as genetic algorithms (GA) and particle swarm optimization (PSO), optimizing the feature set for improved accuracy. Finally, the selected features are classified using four distinct classifiers, achieving remarkable classification accuracies of 99.57% and 98.42% on the COVIDx-2A CT and SARS-CoV-2 CT-Scan datasets, respectively. The novelty of this approach lies in the integration of multiple CNNs and meta-heuristic methods to enhance feature selection and classification performance. Our contributions include the development of a robust diagnostic tool that significantly improves the speed and accuracy of COVID-19 detection, offering a viable alternative to traditional RT-PCR methods.